Integrated Circuits and AI Hardware
Sensors, Actuators and Embedded Memory
R & D Fab
Material, Device, Reliability Analysis
Scanning electron microscope (SEM) image of the in-house fabricated modified LCAT mode resonator based on (a) AlN, and (b) ScAlN. The zoom-in view of the two resonators are depicted below the main images, respectively.
As the RF spectrum gets increasingly crowded, front-end bandpass filters are now becoming the most important element in the applications of wireless communications. Aluminium nitride (AlN) and scandium aluminium nitride (ScAlN) based MEMS resonators are gaining more research interest, with film bulk acoustic resonators (FBARs) and Lamb wave resonators being the two mainstream technology. However, there are current limitations associated with each of these technologies. For example, despite having high quality factor (Q factor) and small size, FBAR which is based on frequency tuning by film stack allows only for single frequency on one wafer which makes it unsuitable for monolithic integration. On the other hand, Lamb wave while based on frequency tuning by lithography which allows for multiple frequency on one wafer, has lower coupling coefficient than FBAR. To overcome the aforementioned challenges, IME is developing a modified laterally coupled alternating thickness (LCAT)-mode resonator based on AlN and ScAlN, which would achieve high coupling coefficient while enabling the capability to achieve monolithic integration.
Heart rate (HR) and breathing rate (BR) monitoring system overview.
Driver fatigue (falling asleep at the wheel) is a worldwide serious problem resulting in many thousands of road accidents each year. It is believed that driver fatigue contributes to more than 30% of road accidents. IME is developing a MEMS accelerometer based sensing solution that non-intrusively monitors parameters such as drivers’ heart and breathing rate for identifying driver fatigue. These data would be transferred wirelessly to a communication system which would alert the driver if in danger of falling asleep behind the wheel. Key features of the sensing solution are small footprint 3-axis MEMS accelerometer, novel combination of MEMS accelerometer and BR/HR detection algorithms, as well as wireless sensors charging via a central power transmission unit which eliminates need for any battery replacement. The system may also be potentially used for other applications such as patient monitoring, sleep quality monitoring and elderly care.
Packaged dew-point sensor.
Dew-point is the temperature at which the rate of water vapour condensation is the same rate as liquid water evaporation. From this measurement, the relative humidity can be calculated. Accurate and precise relative humidity measurements that can work with minimal frequency of calibration over a range of ambient temperatures is crucial for several industrial applications ranging from food processing to materials manufacturing. IME has developed a miniature photonic dew-point sensor, which has high accuracy comparable to commercially available chilled mirror hygrometer, with the added benefits of having a much smaller footprint and manufacturability at a lower cost. In addition, the use of an in situ temperature sensor eliminates temperature errors commonly observed in chilled mirror hygrometers, which are often due to the temperature gradients from the water, mirror and PRT (platinum resistance thermometer) and from self-heating of the PRT.
AlN Energy Harvester
The Internet of Things (IoT) is witnessing a propagation of wireless sensor nodes systems for various sensing objectives. However, there are challenges which remained to be addressed. Some common issues observed include (1) the presence of wired sensors which limits deployment due to accessibility and non-applicability to rotary machinery, (2) battery powered wired systems which limit lifetime and frequency of measurement (for power conservation), hence also restricting sensor deployment to non-critical sites, and (3) the cost of manpower needed to change batteries is prohibitive as number of such systems scales up. There is a need for wireless, self-powered sensor systems to address above issues. Energy harvesting presents a sustainable solution.
IME has developed a vibration based energy harvester which can harness a wide range of vibrations input for harvesting. The energy harvester features a CMOS-compatible design for easy integration with other sensors and ASIC. It has a small footprint to reduce effect of mechanical coupling at critical sites. The overall system is wireless, scalable and flexible to allow for ease of deployment and fuss free integration. A sealed packaged system makes possible deployment in fluid filled environment. The use of such system has potential applications in factory, aerospace, automobile and oil/gas environments.
Analog ReRAM (Non-volatile Memory Devices)
Cross-sectional TEM image of ReRAM chip and memory cell.
Scaling of Moore’s Law has been driving the increased density of traditional memory devices like SRAM, DRAM and non-volatile FLASH. However, with CMOS downsizing beyond 28nm node, issues relating to data retention, higher leakages, low endurance, increased programming and reading voltages set in. With all these challenges, it is difficult to achieve the high bandwidth, memory density and latency required for high performance, throughput and accuracy deep learning-based application (e.g. high-resolution image recognition in autonomous driving). Emerging memory devices like resistive RAM (ReRAM) provides an alternative solution with high memory density, high scalability and low power consumption. Coupled with TSV and 3D integration, data bandwidth can be substantially increased to meet the hardware requirements for high accuracy deep learning computations. In addition, some ReRAM devices possess multi-level states, making it a suitable candidate to realize the synaptic plasticity characteristics observed on biological synapses.
IME provides high yield, high resistance digital and analog ReRAM technology storage solutions suitable for edge computing, deep learning and neuromorphic computing applications. IME’s oxide-based ReRAM devices have been optimized to provide high speed, low power and good reliability. The device stack is highly scalable (for high memory density) and compatible with standard CMOS process flow. Novel 2.5D 1S1R cell array structures are also being pursued for even higher memory density (>2Gbit/cm
Copyright A*STAR Institute of Microelectronics 2019